Design and Implementation Of Cairo University Graphics Processing (CUGPU) Using HLS Approach

Abstract

GPUs can be divided to fixed function GPUs and programmable GPUs. Fixed function GPUs are important for low power mobile applications. This paper addresses the design and implementation, on an FPGA, of a fixed function GPU fully compatible with OpenGL 1.1.12 standard. This GPU is code-named Cairo University GPU (CUGPU). We follow a systematic design approach to design our CUGPU. We developed MATLAB model, C-language model, RTL model and finally a logic model for CUGPU. A hybrid approach for RTL synthesis is followed where automatic and manual synthesis are judiciously combined to achieve small design time and acceptable performance. The final CUGPU is implemented on Xilinx Vertix-7 FPGA. The functionality of CUGPU is verified via direct comparison with the C model results, using hardware simulation and real time on-chip logic analysis.